moby/libnetwork/ipam/allocator.go

411 lines
12 KiB
Go
Raw Normal View History

package ipam
import (
"context"
"fmt"
"net"
"net/netip"
"strings"
"github.com/containerd/log"
"github.com/docker/docker/libnetwork/bitmap"
"github.com/docker/docker/libnetwork/internal/netiputil"
"github.com/docker/docker/libnetwork/ipamapi"
"github.com/docker/docker/libnetwork/ipbits"
"github.com/docker/docker/libnetwork/types"
)
const (
localAddressSpace = "LocalDefault"
globalAddressSpace = "GlobalDefault"
)
// Allocator provides per address space ipv4/ipv6 book keeping
type Allocator struct {
// The address spaces
local, global *addrSpace
}
// NewAllocator returns an instance of libnetwork ipam
func NewAllocator(lcAs, glAs []*net.IPNet) (*Allocator, error) {
var (
a Allocator
err error
)
a.local, err = newAddrSpace(lcAs)
if err != nil {
return nil, fmt.Errorf("could not construct local address space: %w", err)
}
a.global, err = newAddrSpace(glAs)
if err != nil {
return nil, fmt.Errorf("could not construct global address space: %w", err)
}
return &a, nil
}
func newAddrSpace(predefined []*net.IPNet) (*addrSpace, error) {
pdf := make([]netip.Prefix, len(predefined))
for i, n := range predefined {
var ok bool
pdf[i], ok = netiputil.ToPrefix(n)
if !ok {
return nil, fmt.Errorf("network at index %d (%v) is not in canonical form", i, n)
}
}
return &addrSpace{
subnets: map[netip.Prefix]*PoolData{},
predefined: pdf,
}, nil
}
// GetDefaultAddressSpaces returns the local and global default address spaces
func (a *Allocator) GetDefaultAddressSpaces() (string, string, error) {
return localAddressSpace, globalAddressSpace, nil
}
// RequestPool returns an address pool along with its unique id.
// addressSpace must be a valid address space name and must not be the empty string.
// If requestedPool is the empty string then the default predefined pool for addressSpace will be used, otherwise pool must be a valid IP address and length in CIDR notation.
// If requestedSubPool is not empty, it must be a valid IP address and length in CIDR notation which is a sub-range of requestedPool.
// requestedSubPool must be empty if requestedPool is empty.
func (a *Allocator) RequestPool(addressSpace, requestedPool, requestedSubPool string, _ map[string]string, v6 bool) (poolID string, pool *net.IPNet, meta map[string]string, err error) {
log.G(context.TODO()).Debugf("RequestPool(%s, %s, %s, _, %t)", addressSpace, requestedPool, requestedSubPool, v6)
parseErr := func(err error) error {
return types.InternalErrorf("failed to parse pool request for address space %q pool %q subpool %q: %v", addressSpace, requestedPool, requestedSubPool, err)
}
if addressSpace == "" {
return "", nil, nil, parseErr(ipamapi.ErrInvalidAddressSpace)
}
aSpace, err := a.getAddrSpace(addressSpace)
if err != nil {
return "", nil, nil, err
}
if requestedPool == "" && requestedSubPool != "" {
return "", nil, nil, parseErr(ipamapi.ErrInvalidSubPool)
}
k := PoolID{AddressSpace: addressSpace}
if requestedPool == "" {
k.Subnet, err = aSpace.allocatePredefinedPool(v6)
if err != nil {
return "", nil, nil, err
}
return k.String(), netiputil.ToIPNet(k.Subnet), nil, nil
}
if k.Subnet, err = netip.ParsePrefix(requestedPool); err != nil {
return "", nil, nil, parseErr(ipamapi.ErrInvalidPool)
}
if requestedSubPool != "" {
k.ChildSubnet, err = netip.ParsePrefix(requestedSubPool)
if err != nil {
return "", nil, nil, parseErr(ipamapi.ErrInvalidSubPool)
}
}
k.Subnet, k.ChildSubnet = k.Subnet.Masked(), k.ChildSubnet.Masked()
// Prior to https://github.com/moby/moby/pull/44968, libnetwork would happily accept a ChildSubnet with a bigger
// mask than its parent subnet. In such case, it was producing IP addresses based on the parent subnet, and the
// child subnet was not allocated from the address pool. Following condition take care of restoring this behavior
// for networks created before upgrading to v24.0.
if k.ChildSubnet.IsValid() && k.ChildSubnet.Bits() < k.Subnet.Bits() {
k.ChildSubnet = k.Subnet
}
err = aSpace.allocateSubnet(k.Subnet, k.ChildSubnet)
if err != nil {
return "", nil, nil, err
}
return k.String(), netiputil.ToIPNet(k.Subnet), nil, nil
}
// ReleasePool releases the address pool identified by the passed id
func (a *Allocator) ReleasePool(poolID string) error {
log.G(context.TODO()).Debugf("ReleasePool(%s)", poolID)
k, err := PoolIDFromString(poolID)
if err != nil {
return types.InvalidParameterErrorf("invalid pool id: %s", poolID)
}
aSpace, err := a.getAddrSpace(k.AddressSpace)
if err != nil {
return err
}
return aSpace.releaseSubnet(k.Subnet, k.ChildSubnet)
}
// Given the address space, returns the local or global PoolConfig based on whether the
// address space is local or global. AddressSpace locality is registered with IPAM out of band.
func (a *Allocator) getAddrSpace(as string) (*addrSpace, error) {
switch as {
case localAddressSpace:
return a.local, nil
case globalAddressSpace:
return a.global, nil
}
return nil, types.InvalidParameterErrorf("cannot find address space %s", as)
}
func newPoolData(pool netip.Prefix) *PoolData {
ones, bits := pool.Bits(), pool.Addr().BitLen()
numAddresses := uint64(1 << uint(bits-ones))
// Allow /64 subnet
if pool.Addr().Is6() && numAddresses == 0 {
numAddresses--
}
// Generate the new address masks.
h := bitmap.New(numAddresses)
Make the network allocator handle IPv4 blocks too small for network/broadcast addresses. This was originally in docker/libnetwork#2624, which has been closed since the code was moved here. When creating a new network, IPAM's address allocator attempts to reserve the network and broadcast addresses on IPv4 networks of all sizes. For RFC 3021 point-to-point networks (IPv4 /31s), this consumes both available addresses and renders any attempt to allocate an address from the block unsuccessful. This change prevents those reservations from taking place on IPv4 networks having two or fewer addresses (i.e., /31s and /32s) while retaining the existing behavior for larger IPv4 blocks and all IPv6 blocks. In case you're wondering why anyone would allocate /31s: I work for a network service provider. We use a lot of point-to-point networks. This cuts our address space utilization for those by 50%, which makes ARIN happy. This patch modifies the network allocator to recognize when an network is too small for network and broadcast addresses and skip those reservations. There are additional unit tests to make sure the functions involved behave as expected. Try these out: * `docker network create --driver bridge --subnet 10.200.1.0/31 --ip-range 10.200.1.0/31 test-31` * `docker network create --driver bridge --subnet 10.200.1.0/32 --ip-range 10.200.1.0/32 test-32` My installation has been running this patch in production with /31s since March. Signed-off-by: Mark Feit <mfeit@internet2.edu> Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
2021-07-12 21:37:36 +00:00
// Pre-reserve the network address on IPv4 networks large
// enough to have one (i.e., anything bigger than a /31.
if !(pool.Addr().Is4() && numAddresses <= 2) {
Make the network allocator handle IPv4 blocks too small for network/broadcast addresses. This was originally in docker/libnetwork#2624, which has been closed since the code was moved here. When creating a new network, IPAM's address allocator attempts to reserve the network and broadcast addresses on IPv4 networks of all sizes. For RFC 3021 point-to-point networks (IPv4 /31s), this consumes both available addresses and renders any attempt to allocate an address from the block unsuccessful. This change prevents those reservations from taking place on IPv4 networks having two or fewer addresses (i.e., /31s and /32s) while retaining the existing behavior for larger IPv4 blocks and all IPv6 blocks. In case you're wondering why anyone would allocate /31s: I work for a network service provider. We use a lot of point-to-point networks. This cuts our address space utilization for those by 50%, which makes ARIN happy. This patch modifies the network allocator to recognize when an network is too small for network and broadcast addresses and skip those reservations. There are additional unit tests to make sure the functions involved behave as expected. Try these out: * `docker network create --driver bridge --subnet 10.200.1.0/31 --ip-range 10.200.1.0/31 test-31` * `docker network create --driver bridge --subnet 10.200.1.0/32 --ip-range 10.200.1.0/32 test-32` My installation has been running this patch in production with /31s since March. Signed-off-by: Mark Feit <mfeit@internet2.edu> Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
2021-07-12 21:37:36 +00:00
h.Set(0)
}
Make the network allocator handle IPv4 blocks too small for network/broadcast addresses. This was originally in docker/libnetwork#2624, which has been closed since the code was moved here. When creating a new network, IPAM's address allocator attempts to reserve the network and broadcast addresses on IPv4 networks of all sizes. For RFC 3021 point-to-point networks (IPv4 /31s), this consumes both available addresses and renders any attempt to allocate an address from the block unsuccessful. This change prevents those reservations from taking place on IPv4 networks having two or fewer addresses (i.e., /31s and /32s) while retaining the existing behavior for larger IPv4 blocks and all IPv6 blocks. In case you're wondering why anyone would allocate /31s: I work for a network service provider. We use a lot of point-to-point networks. This cuts our address space utilization for those by 50%, which makes ARIN happy. This patch modifies the network allocator to recognize when an network is too small for network and broadcast addresses and skip those reservations. There are additional unit tests to make sure the functions involved behave as expected. Try these out: * `docker network create --driver bridge --subnet 10.200.1.0/31 --ip-range 10.200.1.0/31 test-31` * `docker network create --driver bridge --subnet 10.200.1.0/32 --ip-range 10.200.1.0/32 test-32` My installation has been running this patch in production with /31s since March. Signed-off-by: Mark Feit <mfeit@internet2.edu> Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
2021-07-12 21:37:36 +00:00
// Pre-reserve the broadcast address on IPv4 networks large
// enough to have one (i.e., anything bigger than a /31).
if pool.Addr().Is4() && numAddresses > 2 {
h.Set(numAddresses - 1)
}
return &PoolData{addrs: h, children: map[netip.Prefix]struct{}{}}
}
// getPredefineds returns the predefined subnets for the address space.
//
// It should not be called concurrently with any other method on the addrSpace.
func (aSpace *addrSpace) getPredefineds() []netip.Prefix {
i := aSpace.predefinedStartIndex
// defensive in case the list changed since last update
if i >= len(aSpace.predefined) {
i = 0
}
return append(aSpace.predefined[i:], aSpace.predefined[:i]...)
}
// updatePredefinedStartIndex rotates the predefined subnet list by amt.
//
// It should not be called concurrently with any other method on the addrSpace.
func (aSpace *addrSpace) updatePredefinedStartIndex(amt int) {
i := aSpace.predefinedStartIndex + amt
if i < 0 || i >= len(aSpace.predefined) {
i = 0
}
aSpace.predefinedStartIndex = i
}
func (aSpace *addrSpace) allocatePredefinedPool(ipV6 bool) (netip.Prefix, error) {
aSpace.Lock()
defer aSpace.Unlock()
for i, nw := range aSpace.getPredefineds() {
if ipV6 != nw.Addr().Is6() {
continue
}
// Checks whether pool has already been allocated
if _, ok := aSpace.subnets[nw]; ok {
continue
}
// Shouldn't be necessary, but check prevents IP collisions should
// predefined pools overlap for any reason.
if !aSpace.overlaps(nw) {
aSpace.updatePredefinedStartIndex(i + 1)
err := aSpace.allocateSubnetL(nw, netip.Prefix{})
if err != nil {
return netip.Prefix{}, err
}
return nw, nil
}
}
v := 4
if ipV6 {
v = 6
}
return netip.Prefix{}, types.NotFoundErrorf("could not find an available, non-overlapping IPv%d address pool among the defaults to assign to the network", v)
}
// RequestAddress returns an address from the specified pool ID
func (a *Allocator) RequestAddress(poolID string, prefAddress net.IP, opts map[string]string) (*net.IPNet, map[string]string, error) {
log.G(context.TODO()).Debugf("RequestAddress(%s, %v, %v)", poolID, prefAddress, opts)
k, err := PoolIDFromString(poolID)
if err != nil {
return nil, nil, types.InvalidParameterErrorf("invalid pool id: %s", poolID)
}
aSpace, err := a.getAddrSpace(k.AddressSpace)
if err != nil {
return nil, nil, err
}
var pref netip.Addr
if prefAddress != nil {
var ok bool
pref, ok = netip.AddrFromSlice(prefAddress)
if !ok {
return nil, nil, types.InvalidParameterErrorf("invalid preferred address: %v", prefAddress)
}
}
p, err := aSpace.requestAddress(k.Subnet, k.ChildSubnet, pref.Unmap(), opts)
if err != nil {
return nil, nil, err
}
return &net.IPNet{
IP: p.AsSlice(),
Mask: net.CIDRMask(k.Subnet.Bits(), k.Subnet.Addr().BitLen()),
}, nil, nil
}
func (aSpace *addrSpace) requestAddress(nw, sub netip.Prefix, prefAddress netip.Addr, opts map[string]string) (netip.Addr, error) {
aSpace.Lock()
defer aSpace.Unlock()
p, ok := aSpace.subnets[nw]
if !ok {
return netip.Addr{}, types.NotFoundErrorf("cannot find address pool for poolID:%v/%v", nw, sub)
}
if prefAddress != (netip.Addr{}) && !nw.Contains(prefAddress) {
return netip.Addr{}, ipamapi.ErrIPOutOfRange
}
if sub != (netip.Prefix{}) {
if _, ok := p.children[sub]; !ok {
return netip.Addr{}, types.NotFoundErrorf("cannot find address pool for poolID:%v/%v", nw, sub)
}
}
// In order to request for a serial ip address allocation, callers can pass in the option to request
// IP allocation serially or first available IP in the subnet
serial := opts[ipamapi.AllocSerialPrefix] == "true"
ip, err := getAddress(nw, p.addrs, prefAddress, sub, serial)
if err != nil {
return netip.Addr{}, err
}
return ip, nil
}
// ReleaseAddress releases the address from the specified pool ID
func (a *Allocator) ReleaseAddress(poolID string, address net.IP) error {
log.G(context.TODO()).Debugf("ReleaseAddress(%s, %v)", poolID, address)
k, err := PoolIDFromString(poolID)
if err != nil {
return types.InvalidParameterErrorf("invalid pool id: %s", poolID)
}
aSpace, err := a.getAddrSpace(k.AddressSpace)
if err != nil {
return err
}
addr, ok := netip.AddrFromSlice(address)
if !ok {
return types.InvalidParameterErrorf("invalid address: %v", address)
}
return aSpace.releaseAddress(k.Subnet, k.ChildSubnet, addr.Unmap())
}
func (aSpace *addrSpace) releaseAddress(nw, sub netip.Prefix, address netip.Addr) error {
aSpace.Lock()
defer aSpace.Unlock()
p, ok := aSpace.subnets[nw]
if !ok {
return types.NotFoundErrorf("cannot find address pool for %v/%v", nw, sub)
}
if sub != (netip.Prefix{}) {
if _, ok := p.children[sub]; !ok {
return types.NotFoundErrorf("cannot find address pool for poolID:%v/%v", nw, sub)
}
}
if !address.IsValid() {
return types.InvalidParameterErrorf("invalid address")
}
if !nw.Contains(address) {
return ipamapi.ErrIPOutOfRange
}
defer log.G(context.TODO()).Debugf("Released address Address:%v Sequence:%s", address, p.addrs)
return p.addrs.Unset(netiputil.HostID(address, uint(nw.Bits())))
}
func getAddress(base netip.Prefix, bitmask *bitmap.Bitmap, prefAddress netip.Addr, ipr netip.Prefix, serial bool) (netip.Addr, error) {
var (
ordinal uint64
err error
)
log.G(context.TODO()).Debugf("Request address PoolID:%v %s Serial:%v PrefAddress:%v ", base, bitmask, serial, prefAddress)
if bitmask.Unselected() == 0 {
return netip.Addr{}, ipamapi.ErrNoAvailableIPs
}
if ipr == (netip.Prefix{}) && prefAddress == (netip.Addr{}) {
ordinal, err = bitmask.SetAny(serial)
} else if prefAddress != (netip.Addr{}) {
ordinal = netiputil.HostID(prefAddress, uint(base.Bits()))
err = bitmask.Set(ordinal)
} else {
start, end := netiputil.SubnetRange(base, ipr)
ordinal, err = bitmask.SetAnyInRange(start, end, serial)
}
switch err {
case nil:
// Convert IP ordinal for this subnet into IP address
return ipbits.Add(base.Addr(), ordinal, 0), nil
case bitmap.ErrBitAllocated:
return netip.Addr{}, ipamapi.ErrIPAlreadyAllocated
case bitmap.ErrNoBitAvailable:
return netip.Addr{}, ipamapi.ErrNoAvailableIPs
default:
return netip.Addr{}, err
}
}
// DumpDatabase dumps the internal info
func (a *Allocator) DumpDatabase() string {
aspaces := map[string]*addrSpace{
localAddressSpace: a.local,
globalAddressSpace: a.global,
}
var b strings.Builder
for _, as := range []string{localAddressSpace, globalAddressSpace} {
fmt.Fprintf(&b, "\n### %s\n", as)
b.WriteString(aspaces[as].DumpDatabase())
}
return b.String()
}
func (aSpace *addrSpace) DumpDatabase() string {
aSpace.Lock()
defer aSpace.Unlock()
var b strings.Builder
for k, config := range aSpace.subnets {
fmt.Fprintf(&b, "%v: %v\n", k, config)
fmt.Fprintf(&b, " Bitmap: %v\n", config.addrs)
for k := range config.children {
fmt.Fprintf(&b, " - Subpool: %v\n", k)
}
}
return b.String()
}
// IsBuiltIn returns true for builtin drivers
func (a *Allocator) IsBuiltIn() bool {
return true
}